Research Scientist - Geo-Distributed Inference

Pluralis Research • San Francisco, CA

24d • Remote

About The Position

Pluralis Research is developing Protocol Learning: a way to train and serve large models in a fully decentralized manner on small consumer-grade devices connected over the internet. Its Agora run pretrained an 8B model from scratch on consumer GPUs distributed across the internet, without any participant holding the full weights. The inference pipeline currently generates rollouts for reinforcement learning (RL) training and is intended to serve models after training. It must operate in a permissionless, trustless environment with Macs and consumer GPUs owned by strangers, public-internet connectivity, nodes that may join or leave during a run, and model weights that change as training progresses.

The Research Scientist will build and own the inference stack end-to-end, including pipeline-parallel execution, placement and routing, transport, the serving engine, and failure handling. The role involves inventing and validating methods to make inference fast on consumer hardware over the public internet, then deploying them in production. The scientist will maintain a fast, reliable rollout pipeline for RL training and develop it into the serving layer for trained models.

Pluralis is remote-first, with team members distributed globally and main teams in Australia and North America. The role requires comfort working across time zones and professional-level written and spoken English. The company offers significant equity alongside a high base salary, and optional full visa sponsorship and relocation support to Australia or the United States.

Requirements

• Hands-on experience shipping serving-engine internals or building a large-scale inference system.
• Belief that Protocol Learning is a viable path for collective, trustless, and sovereign AI.
• Professional-level English proficiency, written and spoken.
• Comfort working across time zones.

Nice To Haves

• Publications, including papers or blog posts, in distributed inference or a related field such as LLM serving systems, pipeline parallelism over slow networks, or decentralized training; unpublished work that the candidate can explain is also a strong signal.
• Experience with systems operating in low-bandwidth, high-latency settings such as the public internet.
• Familiarity with RL post-training.
• Exposure to Apple silicon or MLX.
• Experience with P2P networking and NAT traversal.
• Experience at proprietary, open-weight, and open-source AI labs.

Responsibilities

• Build and own the inference stack end-to-end, including pipeline-parallel execution, placement and routing, transport, the serving engine, and failure handling.
• Set the direction for the inference stack and drive its implementation.
• Design, validate, and productionize new methods for fast inference on consumer hardware over the public internet.
• Keep the rollout pipeline fast and reliable for RL training.
• Develop the inference pipeline into the serving layer for trained models.

Benefits

• Significant ownership through an equity-heavy package, in addition to a high base salary.
• Remote-first, flexible work environment with team members distributed globally.
• Optional full visa sponsorship and relocation support to Australia or the United States.
• Opportunity to work on open research problems in decentralized training and serving of frontier models.

What This Job Offers

• Job Type: Full-time
• Career Level: Senior
• Education Level: No Education Listed